Lyra | Medical AI Copilot
Lyra
Try Demo For Clinicians & Teams
Demo data • No real patients

Bring sharper questions to rounds and handoffs.

Lyra turns messy clinical information into concise, evidence‑linked briefs for rounds, handoffs, and escalation—transparent, source‑first, and safe by design.

Try Lyra (Demo Data) Trust & Safety

Example: how Lyra handled a complex case

This is a simulated example of Lyra’s behavior. Every claim shows its source.

Step 1 • Presentation & timeline

68M admitted for dyspnea and edema. PMH: CKD4 (eGFR 23), HTN, T2DM. On furosemide at home.
Timeline:

  • D‑1 18:00 — BMP K⁺ 5.1; ACEi held for AKI.
  • D‑0 20:40 — ACEi restarted per MAR.
  • D‑0 22:00 — Net −600 mL; O₂ sat 96% RA.
  • D‑0 06:10 — BMP K⁺ 6.0; no hemolysis flag.
Vitals
BP 154/88, HR 84, RR 18, SpO₂ 96% RA, afebrile.
Labs
eGFR 23 mL/min/1.73m², BUN 48 mg/dL, Cr 2.5 mg/dL.
Step 2 • Overnight signal detection

• K⁺ increased to 6.0 (↑ from 5.1 yesterday).
• ACE inhibitor restarted at 20:40.
• No repeat BMP ordered since 18:00; last EKG 18h ago.

Lab
BMP 06:10 — K⁺ 6.0 mmol/L; prior 5.1 mmol/L.
Med Order
Lisinopril resumed 20:40 per MAR.
Orders
No repeat BMP order logged; EKG last documented yesterday noon.
Step 3 • Context & pattern

Pattern consistent with renin‑angiotensin blockade–associated hyperkalemia in CKD4. No hemolysis flag; diuresis modest.
Risk context: eGFR 23, diabetes, ACEi exposure — higher risk of true K⁺ elevation.

Guideline
KDIGO CKD/Hyperkalemia: confirm ≥6.0 promptly; consider continuous cardiac monitoring depending on ECG/clinical status.
Chart Note
Problem list: CKD4; outpatient ACEi prior to admission; hemolysis flag absent.
Step 4 • Differential scaffold
  • Don’t‑miss: True hyperkalemia in CKD4 on ACEi; consider telemetry; repeat BMP; review ECG.
  • Less likely: Pseudohyperkalemia — no hemolysis; consistent trend.
  • Contributors to check: K‑sparing diuretics, NSAIDs, acidosis, hemolysis, adrenal insufficiency.
Medication list
No spironolactone/eplerenone; no TMP‑SMX; ACEi active.
ABG/Chem
CO₂ 20 mmol/L (mild acidosis); glucose controlled.
Step 5 • Actions & counterfactuals
  1. Order repeat BMP by 10:00; if ≥6.0 persists, initiate protocol.
  2. Obtain ECG now; monitor telemetry depending on ECG changes/symptoms.
  3. Hold ACEi; review other K‑raising meds; consider loop dose adjustment.

Would change management if… repeat K⁺ ≤5.4 → de‑escalate monitoring; hemolysis flagged → redraw; new peaked T waves → escalate and treat.

Protocol
Service protocol: repeat within 4h when K⁺ ≥6.0 unless hemolysis suspected; treat based on ECG and symptoms.
Pharmacy
No K‑sparing agents; home ACEi resumed; renally dose‑adjusted furosemide under review.
Step 6 • Handover & safety checks

S: 68M CKD4; K⁺ 6.0 after ACEi restart; asymptomatic.

B: K⁺ 5.1 → 6.0 in 24h; no repeat BMP; last ECG 18h ago.

A: True hyperkalemia likely; arrhythmia risk depends on ECG.

R: Repeat BMP by 10:00; ECG now; consider telemetry; hold ACEi; review meds.

  • Safety gates: dose‑check for eGFR 23; alert throttled to ≤0.2/patient‑day; uncertainty shown if data incomplete.
  • Attribution: Draft requires human sign‑off; all claims link to sources.
Audit trail
Reasoning snapshot and source IDs stored for QA; no PHI in logs.
Step 1/6

Three ways Lyra helps today

Watchdog

Context‑aware alerts that default to silence unless confidence and clinical context clear thresholds. Every alert shows why now with source chips.

Rounds Copilot

Evolving problem list, overnight deltas, and don’t‑miss considerations tied to the patient’s history and guidelines. No black boxes—sources inline.

Handover Synthesizer

ISHAPED/SBAR drafts with pending tasks and safety watchpoints. Fully editable, versioned, and attributable to the human signer.

Simple pricing

Day Pass

$9

No login. Demo data + limited tokens.

Start Day Pass

Monthly

$29

Consumer visit‑prep & results explainer modes.

Start Monthly

Clinician

$39

De‑identified cases only. Documentation templates & consideration lists.

For Clinicians & Teams

Try Lyra (Demo Data)

Example vignette (static)

No real PHI. This is a fixed example for illustration only.

68M with CKD4 admitted for volume overload; ACEi restarted last night; this morning BMP shows K+ 6.0. Asymptomatic. Last EKG 18h ago. No repeat BMP ordered since yesterday 18:00. (synthetic)

The real product accepts de‑identified free text. For this page, inputs are fixed so visitors see a clean example immediately.

Visit‑ready brief

Informational only • Review with your clinician

What changed
  • K⁺ 5.1 → 6.0 in 24h; ACEi restarted overnight; no repeat BMP ordered.
Considerations
  • Don’t‑miss: Hyperkalemia in CKD4 — confirm, monitor, review EKG.
  • Probably not: Pseudohyperkalemia (no hemolysis flag; trend consistent).
Actions to confirm
  1. Order repeat BMP by 10:00.
  2. EKG now; adjust QT‑risk meds if present.
  3. Hold ACEi pending potassium control.
SBAR draft

S: CKD4 on ACEi; K⁺ 6.0; asymptomatic.

B: K⁺ rose from 5.1; last EKG 18h ago; no repeat BMP.

A: At risk for arrhythmia; needs confirmation/monitoring.

R: Repeat BMP; EKG now; telemetry; hold ACEi.

Lab
BMP 06:10 — K⁺ 6.0 mmol/L.
Guideline
KDIGO: confirm ≥6.0 promptly; consider cardiac monitoring.
Med Order
ACEi resumed 20:40 per MAR.

Informational synthesis only • Requires human judgment • Not diagnostic or treatment advice.

For Clinicians & Teams

Deployment options

  • On‑prem or private VPC; data residency by region.
  • No PHI in logs; redaction at ingress; role‑based access & audit trails.
  • FHIR/HL7 connectors plus mailbox/SFTP ingestion.

Talk to us

Schedule a 20‑minute discovery call to see Lyra on de‑identified cases and discuss guardrails, governance, and integration scope.

Schedule a Discovery Call

Trust & Safety

Intended Use

Assistive clinical information synthesis and surfacing of contextually relevant considerations for licensed clinicians. Does not diagnose or treat. Requires human review and professional judgment.

Full text
This service provides general medical information and documentation templates for clinical workflow support. It does not deliver patient‑specific treatment recommendations and is not a substitute for professional judgment. For emergencies, call local emergency services.

Model Card (v0)

  • Inputs: de‑identified notes, orders, meds, labs, vitals; guideline snippets.
  • Outputs: summaries, considerations, handoff drafts with citations.
  • Limitations: can miss context outside parsed inputs; defaults to silence when uncertain.
  • Do‑not‑use: life‑threatening triage without clinician; autonomous paging; raw imaging interpretation.

Security Brief

  • Zero‑trust posture; least‑privilege access; network egress controls.
  • At‑rest and in‑transit encryption; rotating keys.
  • Audit‑ready logs with PHI‑minimization; incident response runbooks.

Press & Investors

Fast facts

  • Activation: 63% of visitors generate a brief within 60 seconds (last 14 days, demo sandbox).
  • Helpfulness: 4.4 / 5 average on “Would you bring this to rounds?” (n=187 in‑product ratings).
  • p95 latency: 1.8 s in demo; production target SLA ≥ 99.95% uptime.
  • Gross margin: ~78% at average demo session tokens (static hosting + batched inference).

Methodology: anonymized event telemetry from the public demo; no PHI collected or stored.

Media kit